A Fast and High-Precision Insulator Defect Detection Method

Faults caused by insulator defects in transmission lines can threaten the safe and stable operation of the power grid system, so insulator defect detection plays an important role in transmission line inspection. Currently, the You Only Look Once (YOLO) algorithm in deep learning technology is widely used in insulator fault detection with high accuracy and fast detection. However, insulator defect detection can be affected by bad weather such as foggy days, resulting in a decrease in detection accuracy. In addition, the detection accuracy of the YOLO algorithm itself is crucial to the inspection quality. To solve these problems, this paper proposes an insulator defect detection method. In this method, we use a synthetic fogging algorithm to perform data enhancement on the insulator dataset to solve the problem of degraded detection accuracy in the fog. We introduce coordinate attention block and feature reuse strategy in the YOLOv8 model, and construct an improved model based on YOLOv8, FNYOLO, to realize the high-precision detection of insulator defects. Experiments show that the mean average precision (mAP) of our method is 0.976, and the FPS rate is up to 122, which realizing fast and high-precision insulator defect detection.